Earthquake Analysis For Magnitude 7 And Higher Since 1900
by Wanda Chen
===================
For this project, I would like to answer the following few questions:
-
Which decades had the largest and most earthquakes since 1900?
-
Which area had the most amount of earthquakes since 1900?
-
Is there any relation between magnitude and depth of the epicenter?
-
Which month has the most amount of earthquakes since 1900? (Maybe it related to the gravity of Moon and Sun?)
-
Possible the relationship between magType and magnitude?
Because I was not sure how I should approach the problems, and all I had was the dataset and the graph from the USGS that it created (1900-2013) (http://earthquake.usgs.gov/earthquakes/world/seismicity_maps/), so I may just approach by graphing as many diagrams to search the connection and recreated the graph with wider data range that USGS created. The approach method for this project is “guess -> try -> investigate -> make conclusion(if possible)”.
———————–
Basic Information about the Data
## [1] "time" "latitude" "longitude" "depth" "mag"
## [6] "magType" "nst" "gap" "dmin" "rms"
## [11] "net" "id" "updated" "place" "type"1) table of depth_class 2) summary of magnitude 3) table of mag_class
##
## deep intermediate shallow surface
## 91 126 1090 11## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 7.000 7.100 7.200 7.358 7.600 9.600##
## class1 class2 class3 class4 class5 class6
## 885 341 75 13 3 1category longitude and latitude
Determine the decade for each event -
1) table of decade 2) variable in the dataset 3) information for the variables
##
## 1900s 1910s 1920s 1930s 1940s 1950s 1960s 1970s 1980s 1990 2000s 2010s
## 30 60 112 126 110 101 139 131 110 155 148 96## [1] "time" "latitude" "longitude" "depth" "mag"
## [6] "magType" "nst" "gap" "dmin" "rms"
## [11] "net" "id" "updated" "place" "type"
## [16] "New_Time" "Date" "Time" "Year" "Month"
## [21] "Day" "Hour" "depth_class" "mag_class" "long"
## [26] "long_class" "lat_class" "decade"## 'data.frame': 1318 obs. of 28 variables:
## $ time : Factor w/ 1318 levels "1900-07-29T06:59:00.000Z",..: 546 592 1252 1150 468 1226 605 1272 396 442 ...
## $ latitude : num -38.14 60.91 38.3 3.29 52.62 ...
## $ longitude : num -73.4 -147.3 142.4 96 159.8 ...
## $ depth : num 25 25 29 30 21.6 22.9 30.3 20 15 15 ...
## $ mag : num 9.6 9.3 9 9 8.9 8.8 8.7 8.6 8.6 8.6 ...
## $ magType : Factor w/ 6 levels "","ms","mw","mwb",..: 3 3 6 5 3 5 3 6 3 3 ...
## $ nst : int NA NA 541 601 NA 454 NA 499 NA NA ...
## $ gap : num NA NA 9.5 22 NA 17.8 NA 16.6 NA NA ...
## $ dmin : num NA NA NA NA NA NA NA NA NA NA ...
## $ rms : num NA NA 1.16 1.17 NA 1.09 NA 1.33 NA NA ...
## $ net : Factor w/ 4 levels "atlas","gcmt",..: 3 3 4 4 3 4 3 4 3 3 ...
## $ id : Factor w/ 1318 levels "atlas19230901025800",..: 240 207 1287 1188 322 1261 193 1306 381 338 ...
## $ updated : Factor w/ 1296 levels "2014-02-11T02:25:27.101Z",..: 160 631 1217 1168 40 1218 219 1222 572 549 ...
## $ place : Factor w/ 364 levels "101km SW of Atka, Alaska",..: 52 315 181 236 234 239 270 236 297 89 ...
## $ type : Factor w/ 1 level "earthquake": 1 1 1 1 1 1 1 1 1 1 ...
## $ New_Time : chr "1960-05-22 19:11:20.000" "1964-03-28 03:36:16.000" "2011-03-11 05:46:24.120" "2004-12-26 00:58:53.450" ...
## $ Date : Date, format: "1960-05-22" "1964-03-28" ...
## $ Time : chr "19:11:20" "03:36:16" "05:46:24" "00:58:53" ...
## $ Year : num 1960 1964 2011 2004 1952 ...
## $ Month : num 5 3 3 12 11 2 2 4 4 8 ...
## $ Day : num 22 28 11 26 4 27 4 11 1 15 ...
## $ Hour : num 19 3 5 0 16 6 5 8 12 14 ...
## $ depth_class: chr "shallow" "shallow" "shallow" "shallow" ...
## $ mag_class : Factor w/ 6 levels "class1","class2",..: 6 5 5 5 4 4 4 4 4 4 ...
## $ long : num 287 213 142 96 160 ...
## $ long_class : chr "WestH" "WestH" "EastH" "EastH" ...
## $ lat_class : chr "SouthH" "NorthH" "NorthH" "NorthH" ...
## $ decade : Factor w/ 12 levels "1900s","1910s",..: 7 7 12 11 6 12 7 12 5 6 ...————————–
Univariate Plots Section
Histogram: 1) All data file (7.0 <= mag < 10); 2) 7.0 <= mag < 7.5
## Warning in loop_apply(n, do.ply): Stacking not well defined when ymin != 0
3) magnitude – 7.5 <= mag < 8.0
4) magnitude – 8.0 <= mag < 8.5
5) magnitude > 8.5
Information about the dataset - 1900 - 2015 (April)
Summary of earthquake a) magnitude, b) months, c)depth in the dataset
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 7.000 7.100 7.200 7.358 7.600 9.600## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 1.000 4.000 7.000 6.565 10.000 12.000## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.00 15.00 28.75 71.98 40.00 675.40a) Number of earthquake magnitude and its frequency
b) Number of earthquake happened for each Month
c) Number of earthquake happened for each Hour
##
## 7 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 8 8.1 8.2 8.3 8.4 8.5 8.6 8.7
## 289 205 180 119 92 83 80 77 67 34 21 25 13 13 3 4 6 1
## 8.8 8.9 9 9.3 9.6
## 1 1 2 1 1##
## 1 2 3 4 5 6 7 8 9 10 11 12
## 108 106 109 108 119 98 104 124 90 117 129 106##
## 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
## 55 60 66 63 51 45 54 50 53 50 61 50 49 48 68 45 58 53 57 62 51 62 65 42Information about amount of epicenter location
d) Number of earthquake happened in East hemisphere and West hemisphere
e) Number of earthquake happened in North hemisphere and South hemisphere
##
## EastH WestH
## 900 418##
## NorthH SouthH
## 720 598Histogram of the depth of epic center (km)
number distribution of each depth
Other plots that look at relation of magnitude, depth, occurance frequency and variance of earthquakes.
## Warning in loop_apply(n, do.ply): Removed 2 rows containing missing values
## (geom_path).## Warning in loop_apply(n, do.ply): Removed 2 rows containing missing values
## (geom_path).## Warning in loop_apply(n, do.ply): Removed 2 rows containing missing values
## (geom_path).## Warning in loop_apply(n, do.ply): Removed 2 rows containing missing values
## (geom_path).## Warning in loop_apply(n, do.ply): Removed 2 rows containing missing values
## (geom_path).## Warning in loop_apply(n, do.ply): Removed 2 rows containing missing values
## (geom_path).
Number of earthquakes for each decade
##
## 1900s 1910s 1920s 1930s 1940s 1950s 1960s 1970s 1980s 1990 2000s 2010s
## 30 60 112 126 110 101 139 131 110 155 148 96———————–
Univariate Analysis
What is the structure of your dataset?
This is the basic structure of the .csv file that I downloaded:
time - Time when the event occurred. Times are reported in milliseconds since the epoch ( 1970-01-01T00:00:00.000Z), and do not include leap seconds. In certain output formats, the date is formatted for readability.
latitude - Decimal degrees latitude. Positive values for northern latitudes. Negative values for southern latitudes.
longitude - Decimal degrees longitude. Positive values for eastern longitudes. Negative values for western longitudes.
depth - depth of the event in the kilometers
mag - The magnitude for the event
magType - The method or algorithm used to calculate the preferred magnitude for the event. Includes “Md”, “Ml”, “Ms”, “Mw”, “Me”, “Mi”, “Mb”, “MLg”.
nst - The total number of Number of seismic stations which reported P- and S-arrival times for this earthquake.
gap - The largest azimuthal gap between azimuthally adjacent stations (in degrees). In general, the smaller this number, the more reliable is the calculated horizontal position of the earthquake.
dmin - Horizontal distance from the epicenter to the nearest station (in degrees). 1 degree is approximately 111.2 kilometers. In general, the smaller this number, the more reliable is the calculated depth of the earthquake.
rms - The root-mean-square (RMS) travel time residual, in sec, using all weights. This parameter provides a measure of the fit of the observed arrival times to the predicted arrival times for this location. Smaller numbers reflect a better fit of the data. The value is dependent on the accuracy of the velocity model used to compute the earthquake location, the quality weights assigned to the arrival time data, and the procedure used to locate the earthquake.
net - The ID of a data contributor. Identifies the network considered to be the preferred source of information for this event. The value includes: ak, at, ci, hv, ld, mb, nc, nm, nn, pr, pt, se, us, uu, uw.
id - A (generally) two-character network identifier with a (generally) eight-character network-assigned code.
updated - Time when the event was most recently updated.
place - Textual description of named geographic region near to the event. This may be a city name, or a Flinn-Engdahl Region name.
The variables that were created from the dataset -
Date - the date when the event happened
Year - the year when the event happened (4 digits value; 1900 - 2015)
Month - the month when the event happened (1-2 digits value; 1-12)
Day - the day when the event happened (1-2 digits value; 1-31)
Hour - the hour when the event happened (1-2 digits value; 0-23)
depth_class - different zones for the depth of the epicenter:
depth = 0 – character string - “surface”
0 < depth < 70 – character string - “shallow”
70 <= depth <= 300 – character string - “intermediate”
depth > 300 – character string - “deep”
The zones (except depth=0) is according to USGS defined
long - it is the another way to express of longitude that shows center of graph is 180 instead of 0
mag_class - the class that shows the group for the magnitude:
class 1 - magnitude < 7.5
class 2 - 7.5 <= magnitude < 8.0
class 3 - 8.0 <= magnitude < 8.5
class 4 - 8.5 <= magnitude < 9.0
class 5 - magnitude >= 9.0
long_class - the class that shows the group for the longitude
WestH - longitude > 180.0
PrimeMeridian - longitude = 180.0
EastH - longitude < 180.0
lat_class - the class that shows the group for the latitude
NorthH - latitude > 90.0
Equator - latitude = 90.0
SouthH - latitude < 90.0
Decade - for each 10 years in the Year field, it will determine which decade the time is in. It ranges from the 1900s all the way to the 2010s.
What is/are the main feature(s) of interest in your dataset?
I am also interested in the month of the earthquake, and I want to know if the Sun, Moon, and the Earth, may have a relationship to earthquakes.
What other features in the dataset do you think will help support your investigation into your feature(s) of interest?
magType field may help in my investigation. Unfortunately, there is no fault line information; I think that fault line information will help even more than just the magType.
Did you create any new variables from existing variables in the dataset?
Yes.
The time field in the dataset includes date and time. I am interested in the Year and Month fields, so I split the time field into Date, and Time. The Date field is split into 3 fields - Year, Month, Day. I only extracted the Hour from the Time field.
depth_class - a field that shows different depth class (surface, shallow intermediate, deep) according to its depth.
mag_class - a field that shows different magnitude class (separated by range of 0.5 magnitude)
long_class - a field that describes the longitude on the earth (180 degree as divider). West Hemisphere, East Hemisphere, Prime Meridan.
lat_class - similar to long_class, but it describes latitude (0 degree is the divider). North Hemisphere, South Hemisphere, Equator.
decade - it shows the number of the decade for the year.
Of the features you investigated, were there any unusual distributions? Did you perform any operations on the data to tidy, adjust, or change the form of the data? If so, why did you do this?
I adjusted the dateset to make sure the date is split off from the time field. The Date field splits to 3 fields - Year, Month, Day, so it can help me to analyze the data set.
From the data it is rare to have magnitude 9 or higher earthquakes.
Also it is less likely to have large earthquakes that are magnitude 8.5 or higher and depth is more than 100 km.
———————–
Bivariate Plots Section
(1) For this section, I would like to investigate to see if there is any relation among any two variables.
Want to look though location of epicenter from different point of view
scatterplot to show where earthquakes happened
1) depth of the epicenter - with the original latitude and longitude from the given dataset
2) Hours distribution - with modified longitude and original lattitude
3&4) mag_class distribution - with modified longitude and original lattitude
5&6) depth of the epicenter - with the original latitude and modified longitude
1) Most of earthquakes happen around on the border of the Pacific Ocean except south border. 2) It is not often to have deep (depth >= 300km) earthquake, shallow (depth < 30 km) earthquakes happen most often. 3) Most of earthquakes are class1 (7.0 <= magnitude <7.5) - class3 (magnitude < 8.5).
(2) Other possible variables summary
Just want to know if other variables may produce interesting result
1) magnitude methods/algorithms 2) network distribtion
## ms mw mwb mwc mww
## 7 94 907 46 193 71
## atlas gcmt iscgem us
## 7 1 731 579Most of earthquake magType (the method or algorithm used to calculate the preferred magnitude) is “mw” then “mwc”. The top 2 data contributor are: “iscgem” and “us”.
(3) Investigate the data that is difference among the Richter magnitude scale
Richter magnitude scale was defined in 1935, and widely used after 1970. The graphs here shows how the earthquake spread out: 1)before 1935; 2) Between 1935 and 1970; 3)After 1970.
Although Richter magnitude scale did not widely use until after 1970, most of the earthquake stil happened around “Ring of Fire”, no matter which period (before 1935, between 1935 and 1970, or after 1970). Also it happend more on the western side of ring than eastern side.
(4) Epicenter analysis -
Want to look through different depth, magnitude, and time relation to see if there were some relations exist among any of them.
1) magnitude vs depth_class w/ linear regression line
2) depth vs mag_class w/ linear regression line
3) depth vs Month w/ linear regression line
4) Month vs depth w/ mean line
5) Month vs magnitude w/ linear regression line
6) Year vs magnitude w/ mean line
## Warning in loop_apply(n, do.ply): Removed 150 rows containing missing
## values (geom_point).
## Warning in loop_apply(n, do.ply): Removed 5 rows containing missing values
## (geom_point).
## Warning in loop_apply(n, do.ply): Removed 1 rows containing missing values
## (stat_smooth).## Warning in loop_apply(n, do.ply): Removed 115 rows containing missing
## values (geom_point).
## Warning in loop_apply(n, do.ply): Removed 1 rows containing missing values
## (stat_summary).## Warning in loop_apply(n, do.ply): Removed 109 rows containing missing
## values (geom_point).
## Warning in loop_apply(n, do.ply): Removed 247 rows containing missing
## values (geom_point).
## Warning in loop_apply(n, do.ply): Removed 145 rows containing missing
## values (geom_point).
When the graph is compared with magnitude, most of linear regression line is less than 7.5 Richter magnitude scale. When the graph is compared with depth, the linear regression line is along 75 km line.
(5) Boxplot analysis - median value for each mag_class
I want to see how the median value for each relation comparison
1) mag_lass vs Year - class1 -> class 5: 1970, 1965, 1958, 1960, 1984
2) mag_class vs Month - class1 -> class 5: 6, 7, 7, 4, 4
3) mag_class vs Hour - class1 -> class5: 12, 11, 12, 14, 4
4) Hour vs magnitude
5) Month vs magnitude
6) decade vs magnitude
The median range for class1 through class4 (magnitude < 9.0) is on 1960s, except class 5 (magnitude >= 9.0), which is 1980s. The median range for Month to have earthquakes in between April through July. And the median range for earthquake to happen is UTC hour 11 through hour 14, except class 5 that median is hour 4. The median magnitude for decade, Month, and Hour is around 7.2 to 7.4.
(6) Magnitude and Depth Relationship
I want to see when it is at different % level, will it make better outcome?
1) magnitude and depth relationship @ 95%
both linear regression and mean line is along 50 km line
2) magnitude and depth relationship @ 99%
both linear regression and mean line is along 75 km line
## Warning in loop_apply(n, do.ply): Removed 110 rows containing missing
## values (stat_smooth).## Warning in loop_apply(n, do.ply): Removed 110 rows containing missing
## values (stat_summary).## Warning in loop_apply(n, do.ply): Removed 110 rows containing missing
## values (geom_point).
## Warning in loop_apply(n, do.ply): Removed 27 rows containing missing
## values (stat_summary).## Warning in loop_apply(n, do.ply): Removed 27 rows containing missing
## values (stat_smooth).## Warning in loop_apply(n, do.ply): Removed 27 rows containing missing
## values (geom_point).
———————–
Bivariate Analysis
Talk about some of the relationships you observed in this part of the investigation. How did the feature(s) of interest vary with other features in the dataset?
From the graphs that I picked, it shows a strong relationship among magnitudes and the locations of the epicenter. A lot of major ones are located in the “Ring Of Fire” area, which is on either side of the Pacific Ocean.
It also shows strong earthquakes located in Southeast Asia, and the China/India border. It shows similarly in the graph that USGS created that most of the earthquakes happen in the border of plate boundaries “http://earthquake.usgs.gov/earthquakes/world/seismicity_maps/world.pdf”.
I thought the magnitude and depth of the epicenter has some kind of correlation, but when I ran the test, all the values are really close to 0. So, it shows there is no direct relation between the depth and magnitude. It might have some relation if the data set included casualty and financial lost of the earthquakes. However, with large earthquakes (magnitude >=8.5), the depth does have a small correlation with the magnitude; it usually happens with shallow or surface area compared with intermediate or deep earthquakes.
Did you observe any interesting relationships between the other features (not the main feature(s) of interest)?
One of the things may be just coincidence: the earthquakes that happened in April are all shallow class so far; the earthquakes that happened in October were shallow and deep class; and the earthquakes that happened in December were shallow and intermediate. The rest of the months included all types of earthquakes. Magnitudes that are 9 and above happened in March, May, and December.
One of the methods of algorithm used to calculate the preferred magnitude for the earthquakes is “mw” including a few of the largest earthquakes, which its net (data contributor) is “iscgem”.
One of graphs that I found interesting is the earthquake magnitude class vs. the year it happened. The boxplot shows that the median of each mag_class is all around 1960’s except for magnitude >= 9.0 (class5), and the result for the mean is also similar to the median. The graph for Month vs mag_class also shows similar results. Both mean and median for magnitude < 8.5 (class1, class2, class3) is around June and July; but for magnitude >= 8.5 (class4 and higher) the median is on April, and Mean is somewhere in May. With mag_class vs Hour, for class1 - class4, the mean is around Hour 11 - Hour 12. For median it is also around Hour 11 - Hour 12 for class1 - class3, but the median for class4 is Hour 14, which is different than its mean.
What was the strongest relationship you found?
The strongest relationship I found is the location (latitude and longitude) vs magnitude of the earthquakes. It sits into the “Ring of Fire”, which is from southeast of the Pacific Ocean toward north and all the way to the southwest of Pacific Ocean. Also even though the Richter magnitude scale was defined around 1935, it did not become widely used until after 1970. All the earthquakes through out the 3 different time still show the Ring of Fire has the most earthquakes than other area.
———————–
Multivariate Plots Section
(1) More detail analysis on magnitude and depth of epicenter and its occurrance
From bivariate plots, but added in another variable to see if I can make better analyze the data.
a) Year vs magnitude through depth_class
b) location of earthquake distribution - mag_class
c) location of earthquake distribution - depth point of view
The second and third graphs helped me achieved one of the goal that I would like to complete for this project, which is recreated the graph that I got from the USGS website. Although the two graphs are separated, it does show many similar relations. Most of the earthquakes are around the Ring of Fire. It is less likely to have great magnitude and epicenter is in deep class earthquakes. Shallow and intermediate class earthquakes still occur more than deep.
(2) Investigate magnitude and depth of Epicenter (km)
Want to know if there is correlation through the comparison of magnitude and depth
a) scatterplot between the magnitude and depth of epoch center (km)
The mean for each magnitude is around or less than 100 km deep; linear regression line is between 50 km and 100 km.
b) Magnitude vs Epi Center Depth < 70 km
Surface and shallow epi center the mean is between 20 km and 30 km; linear regression line is around 25 km line.
c) Magnitude vs Epoch Center Depth between 70km and 300km
The mean of depth is between 100 km and 180 km, except magnitude = 8.1 which its mean depth is 240; linear regression line is from 150km at magnitude 7.0 downward to 135km at magnitude 8.3.
d) Magnitude vs Epoch Center Depth >= 400 km
The mean epi center is around 520 km and 560 km, except magnitude 7.4, 8.0-8.3, the depth are 460 km, and over 600 km.
##
## Pearson's product-moment correlation
##
## data: mag and depth
## t = -1.1972, df = 1316, p-value = 0.2315
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.08682423 0.02105088
## sample estimates:
## cor
## -0.03298273
##
## Pearson's product-moment correlation
##
## data: mag and depth
## t = -0.9579, df = 1099, p-value = 0.3383
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.08781253 0.03024934
## sample estimates:
## cor
## -0.02888232
##
## Pearson's product-moment correlation
##
## data: mag and depth
## t = -1.1972, df = 1316, p-value = 0.2315
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.08682423 0.02105088
## sample estimates:
## cor
## -0.03298273
##
## Pearson's product-moment correlation
##
## data: mag and depth
## t = 0.6564, df = 89, p-value = 0.5133
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.1385166 0.2714726
## sample estimates:
## cor
## 0.069408231) Most of earthquakes that is magnitude > 8.5, they are shallow earthquakes. 2) The mean depth for each depth class is: a) surface and shallow class (depth < 70 km), the epi center is between 20km and 30 km b) intermediate class (70km <= depth < 300km) is around 100 km and 180 km c) the depth class (depth >= 300 km), most of the mean epi center is around 520 km and 560 km, but there are more exceptions in the depth class. 3) All those correlation are all closed to 0 which means it does not have direct relationship.
(3) How earthquakes spread throught out the years and months
I want to see how for the past 115 years, how earthquakes had spread out throught out each Month
Year vs Month graphs:
i) from the point of depth_class
ii) from the point of magnitude_class
## Warning in loop_apply(n, do.ply): Removed 129 rows containing missing
## values (geom_point).
## Warning in loop_apply(n, do.ply): Removed 109 rows containing missing
## values (geom_point).
It looks there were less reported earthquakes before 1920s, and it looks like there were less earthquakes in 1950s (1950-1959). Most of earthquake still in class1 and class 2, and shallow class epicenter. It shows that large earthquakes happened all the decades and any month in the year. The linear regression line is between the month of June and July. No direct relations among earthquakes and time of years and months.
(4) Occurance of earthquake through different combination view of mag_class, depth_class, longitude, latitude, magnitude, and depth.
Again, I want to see if the occurrances of earthquakes can help to make more sense on magnitude-depth-location relation.
## Warning in loop_apply(n, do.ply): Removed 6 rows containing missing values
## (geom_point).
Clearly that for class4 and higher, they were all shallow earthquakes. Even for class3 it rarely had deep class earthquakes and less likely to have intermediate class earthquakes. When I looked at depth vs longitude and depth vs latitude graphs how epicenter spread out through each magnitude class, the distribution of class1 (magnitude < 7.5) and class2 (78.5 <= magnitude < 8) is similar.
(5) Look at how earthquakes spread out for each hemisphere by investigated a)magnitude, b)depth
I saw the big picture how earthquakes spread through the all longitudes and latitudes display. Now I want to focus on how earthquakes spread out throught different hemisphere.
1) East hemisphere vs West hemisphere
2) North hemisphere vs South hemisphere
Most of East hemisphere earthquake is above -30 degree latitude area (-30 to 60); West hemisphere earthquakes are more widely spread, but it shows more around -40 to 20 degree latitude area.
Most southern hemisphere earthquakes were focused around 120 to 190 degree and 280 to 300 degree longitude area; the concentrated area was 150 to 190 degree, and 280 and 300 degree area. Northern hemisphere’s earthquakes were widely spread across all longitudes area, but it showed concentration around 120 and 165 degree. For southern hemisphere, it showed less likely to have earthquakes that were between 0 and 30 degree longitude and 195 and 270 degree longitude.
(6) Scatterplot Matrix Analysis for each hemisphere
When I investigated all the individual, bivariate, and multivariate graphs, I can see if there is any direct relation among the variables. When I want to see the big picture for comparison, scatterplot matrix seemed a good fit to see the relation.
Each matrix show each hemisphere analysis for different fields. For each hemisphere, the most significant relation are all the line graphs on the diagonal line that can be put on top of certain scatterplots. For example: latitude-depth, longitude-depth, depth-magnitude.
It is really hard to find direct relation between latitude and other variables. But when I looked at different hemispheres for longitude and depth, it seemed I can put the line graph directly on (long, depth) and (depth, mag) graph directly. Although all the correlation values are closer to zero, the following correlation values are farther away from zero: the south hemisphere [(long, latitude), -0.491], west hemisphere [(long, latitude), -0.264], and east hemisphere [(long, latitude), -0.327]. South hemisphere’s correlation shows the strongest among all hemispheres, and it means if there is an earthquake in the south hemisphere, it is more likely to have 50% chance to predict correctly if either longitude or latitude is known.
(7) Mean/median analysis - Hour, Month, and Year/Decade
I want to see if using conditional means and grouping can help analyze the data on the time variables
(i) Investigate Hour with Magnitude Mean/Median
## Warning in loop_apply(n, do.ply): Removed 193 rows containing non-finite
## values (stat_boxplot).## Warning in loop_apply(n, do.ply): Removed 193 rows containing missing
## values (stat_summary).
When used conditional mean, the hours_mean was around magnitude 7.3 and 7.4, except for hour 5 and hour 7; and without it, the mean range was around magnitude 7.2 and 7.3.
(ii) Investigate Month with Magnitude Mean/Median
## Warning in loop_apply(n, do.ply): Removed 126 rows containing non-finite
## values (stat_boxplot).## Warning in loop_apply(n, do.ply): Removed 126 rows containing missing
## values (stat_summary).
Again, when used conditional mean, the month_mean was around magnitude 7.3 and 7.4; and without it, the mean range was around magnitude 7.2 and 7.3.
(iii) Investigate decade with Magnitude Mean/Median
When used conditional mean, the decade_mean was around magnitude 7.3 and 7.4, except for decades: 1910s, 1920s, and 1980s; and without it, the mean range was around magnitude 7.2 and 7.3, except 1910s and 1920s.
(8) Focus on the [80<longitude<300,-70<latitude<75] location of the earthquake only
Since the earthquakes spreaded out widely around the world, now I want to focus on the area along the “Ring of Fire”. Again, I will investigate through different hemisphere in the given range, from mag_class, depth_class, and time perspective.
## Warning in loop_apply(n, do.ply): Removed 942 rows containing missing
## values (geom_point).
## Warning in loop_apply(n, do.ply): Removed 942 rows containing missing
## values (geom_point).
## Warning in loop_apply(n, do.ply): Removed 513 rows containing missing
## values (geom_point).
## Warning in loop_apply(n, do.ply): Removed 513 rows containing missing
## values (geom_point).
After narrow down the area, it showed certain area had earthquakes more than the other.
(9) Focus even more smaller area
I want to see if I can have more detail look through smaller and condense area in the “Ring of Fire”.
East hemisphere - Focus on 90 <= longitude <=170, -30<=latitude <=60
West hemisphere - Focus on 255 <= longitude <=300, -55<=latitude <=20
## Warning in loop_apply(n, do.ply): Removed 1098 rows containing missing
## values (geom_point).
## Warning in loop_apply(n, do.ply): Removed 1098 rows containing missing
## values (geom_point).
## Warning in loop_apply(n, do.ply): Removed 600 rows containing missing
## values (geom_point).
## Warning in loop_apply(n, do.ply): Removed 600 rows containing missing
## values (geom_point).
After narrow down even more, it showed more clear how epicenter distributed. Eastern hemisphere showed lots more epicenter than western hemisphere.
(10) Focus on each quadrant
Previous several graphs, it focus on smaller area or different hemispheres. Now I want to see how epicenter distributed across from each quadrant.
Each quadrant will include: i) from mag_class; ii) from depth_class point of view.
(a) Q1 - 180 < longitude < 360 & 0 < latitude < 90
(b) Q2 - 0 < longitude < 180 & 0 < latitude < 90
(c) Q3 - 0 < longitude < 180 & -90 < latitude < 0
(d) Q4 - 180 < longitude < 360 & -90 < latitude < 0
## Warning in loop_apply(n, do.ply): Removed 8 rows containing missing values
## (geom_path).
## Warning in loop_apply(n, do.ply): Removed 8 rows containing missing values
## (geom_path).
##
## Pearson's product-moment correlation
##
## data: mag and depth
## t = 0.4643, df = 187, p-value = 0.643
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.1093241 0.1758144
## sample estimates:
## cor
## 0.03393572
##
## Pearson's product-moment correlation
##
## data: mag and depth
## t = -0.2816, df = 529, p-value = 0.7783
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.09723304 0.07292208
## sample estimates:
## cor
## -0.01224412## Warning in loop_apply(n, do.ply): Removed 17 rows containing missing
## values (geom_path).
## Warning in loop_apply(n, do.ply): Removed 17 rows containing missing
## values (geom_path).
##
## Pearson's product-moment correlation
##
## data: mag and depth
## t = -1.1841, df = 367, p-value = 0.2372
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.16275713 0.04065794
## sample estimates:
## cor
## -0.06169016
##
## Pearson's product-moment correlation
##
## data: mag and depth
## t = -1.0876, df = 227, p-value = 0.2779
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.19977833 0.05818228
## sample estimates:
## cor
## -0.07200198For Q1 (North hemisphere and West hemisphere), the linear regression line is negative (latitude 55 degree down to 0 degree). For Q2 (North hemisphere and East hemisphere), the linear regression line is positive (latitude - 20 degree to 35 degree). For Q3 (South hemisphere and East hemisphere), the linear regression line is negative (latitude - 0 degree down to 18 degree)For Q4 (South hemisphere and West hemisphere), the linear regression is negative (latitude - -20 degree to -35 degree). Q2 is the only quadrant that has positive linear model line.
————————————
Multivariate Analysis
Talk about some of the relationships you observed in this part of the investigation. Were there features that strengthened each other in terms of looking at your feature(s) of interest?
From the bivariate analysis, I already know that most of earthquakes happened around the “Ring of Fire”. here, I separated the latitude and longitude, and I wanted to know if certain latitude or longitude happened more than others. From the depth of the epicenter, the longitude between 90 and 190, and 260 and 300, there were lots of occurrences of earthquakes. Among them, longitude between 100 and 180, and 280 and 300 had more deep earthquakes. When I checked the map of the earth, the longitude is the international dateline, and it is located in the middle of Pacific Ocean.
Were there any interesting or surprising interactions between features?
The distribution of the earthquakes for magnitude class and depth class are similar. Although most of the earthquakes are either shallow or intermediate, there were a few exceptions with magnitude around 7, but whose depth belong to “deep” class. It looks like Europe or Africa do not have a lot of large earthquakes (magnitude 7 and higher) compared to Asia and America. The exception that happened a lot and does not belong to the “Ring of Fire” is Afghanistan, and the area that goes through border of Europe. North hemisphere has more large earthquakes than south hemisphere, and east hemisphere has more than twice that of west hemisphere.
OPTIONAL: Did you create any models with your dataset? Discuss the strengths and limitations of your model.
————————————
Final Plots and Summary
Plot One
Description One
It shows most of earquakes around the Ring of Fire. Most class 1 earthquakes (magnitude > 7.5) are shallow earthquakes. Also for magnitude >= 7.0 earthquakes, high magnitude earthquakes are less likely to happen. It is less likely to have magnitude more than 8.0. It shows that the right side of the “Ring of Fire” has fewer deep eathquakes than the left side of the ring. And most deep earthquakes are around the ocean area especially on the left side of the ring.
It is less likely to have large earthquakes in Europe or Africa areas. Continent Asia had more earthquakes than North and South America.
From this graph, we can also tell that the area around the Pacific Ocean has more earthquakes than the Indian or Atlantic Ocean.
Most of the deep class earthquakes are in the longitude range 100 to 180 degree and 280 to 300 degree. It’s not like the shallow or intermdiate class that are widely spread throughout all longitudes.
It is less likely to have magnitude 7 or higher earthquakes that belong to intermediate or deep class inside of Arctic Circle (66°33′45.7″ north of the Equator) or Antarctic Circle (66°33′45.8″ south of the Equator).
————————————
Plot Two
## Warning in loop_apply(n, do.ply): Removed 126 rows containing non-finite
## values (stat_boxplot).## Warning in loop_apply(n, do.ply): Removed 126 rows containing missing
## values (stat_summary).## Warning in loop_apply(n, do.ply): Removed 126 rows containing non-finite
## values (stat_boxplot).## Warning in loop_apply(n, do.ply): Removed 126 rows containing missing
## values (stat_summary).## Warning in loop_apply(n, do.ply): Removed 126 rows containing non-finite
## values (stat_boxplot).## Warning in loop_apply(n, do.ply): Removed 126 rows containing missing
## values (stat_summary).
Description Two
These 3 graphs helped me solve the following questions that I had when I started this project for earthquakes of magnitude 7 or higher since 1900:
- Which decade has the largest earthquake? - 1960s (9.6)
- Which decade has the most earthquake? - 1990s (155)
- Which decade has the highest median earthquake? - 1900 (7.70)
- Which decade has the highest mean earthquake? - 1900 (7.70)
- Which month has the highest mean earthquake? - December (7.414)
- Which month has the highest median earthquake? - September (7.4)
- Which month has the most amount of earthquake? - November (129)
- Which month has the least amount of earthquake? - September (90)
- Which hour has the highest median erthquake? - Hour 23 (7.4)
- Which hour has the highest mean erthquake? - Hour 5 (7.46)
- Which hour has the most amount of earthquake? - Hour 14 (68)
- Which hour has the least amount of earthquake? - Hour 23 (42)
- When look through the mean and median comparison, most of mean is above or on the median line for decade, Month, and hour graphs, except 1900s and Hour 23. On the Month of May, its median is also same as its first quartile.
- Since most of the third quartile is less than magnitude 7.6, so I can conclude that there is less than 25% chance to have magnitude 8.0 or higher earthquakes anytime. According to the Wikipedia, for magnitude 8.0 - 8.9, it estimated once per year, and 9.0 and higer, one per 10 to 50 years, and no like magnitude 7.0 - 7.9, it estimated average 10 to 20 per year.
————————————
Plot Three
Description Three
- When looking at the correlation values, it seems none of the fields are related to each other since the correlation values are almost 0.
- When I look at the (latitude, long) graph, it looks like the ring of fire in its reverse side way, which is similar to plot 1 that shows most of the earthquakes area.
- When I look at (long, depth) and (long, mag) graphs, it looks like I can put the curve line on top of those 2 graphs and shows a few outliers on longitude is 200 and depth is more than 400. It also shows that a lot of magnitude between 7 and 8 earthquakes happened in the longitude between the 100 to 200 degree area.
- Depth column shows most of earthquakes belong to shallow class, even through out all magnitudes.
—————————
# Reflection
-
When I started this project, these are some of questions that I want to answer:
- Which decades had the largest and most earthquakes since 1900?
- Which area had the most amount of earthquakes since 1900?
- Is there any relation between magnitude and depth of the epicenter?
- Which month has the most amount of earthquakes since 1900? (Maybe it related to the gravity of Moon and Sun?)
- Possible the relationship between magType and magnitude? I wasn’t even know if I can draw any conclusion at all. Because people cannot really control when and where earthquake to happen, also it is hard to tell that if there is a large earthquake in Japen, and few days later, there is aother large earthquake in China or in the Java Sea. We cannot really conclude those are somewhat indirectly related to each other.
Where did I run into difficulties in the analysis?
There were many difficulties for this project (before, during and after):
- Before I started this project, I wasn’t sure what kind of project I would do. It took me over a week to decide and search for a dataset.
-
When I decided to do this project, I was just interested in the area of earthquakes, and how many large earthquakes has happened so far. I never really thought that I would be able to draw any conclusions from the dataset. I also wasn’t sure:
- how large is considered a large earthquakes.
- what is the year range I should consider.
- what kinds of questions I should find out
- Will I be able to find out the answeres that I want to find.
- There are only few numerical fields and few categorical fields that are able to do the analysis. Many fields that I created were after some struggles and not sure how I should do, than I created to help me to view on many occassions.
- Because each earthquake is independent from each other (unless there is a field of data shows it was aftershock of the major one), it is really difficult to create some graphs, such as line or bar graphs. It seems scatter plot was the best choice. Also it is a discrete event for all the data, so it just does not make sense to find the average, mean, or quartiles.
- After creating so many graphs, I wasn’t sure what kind of analysis or conclusion I can make out of these. Basic data information were created by USGS, I can’t really change some of the information. Some fields looked interesting and I thought I might be able to use them to run analysis out of it, but it turned out many data for those variables have no information.
- There is lots of repeated code because I want to make sure when I compared each group, they have same basic template to compare though using different variables.
- I cannot really conclude or make positive analysis that my analysis will be correct for future earthquakes prediction. Because each fault line, or the reasons that cause earthquakes, are unknown, in this dataset, the Earth has a mind of its own.
Where did I find successes?
The first success that I found was when I created the plot 1 graph that is similar to the one USGS created, which it also helped me solve many of the difficulties/struggles that I had when I worked on this project. Although I do not include the map, plate bouundaries, and active volcanoes locations, I feel I did the best as I can. It proves that most of the earthquakes happen around “the Ring of Fire”.
Through out this project, it helped me understand more about how to create different graphs using ggplot. Some of graphs in this datasets just not suitable than the other.
When I found out that most of earthquakes are shallow, and are rarely deep earthquakes, that kind of explains to me why many earthquakes cause severe damage/casualties when we hear that the eathquake is magnitude 7 or higher and it is shallow either from the news or other datasets. Also throughout out the years, every year will have at least 1 large earthquake that is magnitude 7 or higher.
And for the past 100+ years, there is no earthquake that is magnitude 7 or higher recording that happened at either Noth Pole or South Pole.
How could the analysis be enriched in future work?
I got this dataset through USGS website. Unfortunately in this data collection, it does not include some important information for this dataset. If it included:
- faultline - then maybe we can conclude or have a better prediction which faultline has a higher probability to have large earthquakes.
- Phase of the Moon - we might be able to determine if it has some kind of relationship with the occurence of earthquakes.
- Location of volcano - it can help to decide if because of volcanic activity that might affect the earthquake/faultline to occur.